Create and manage instance groups with Kubernetes.
instance-manager simplifies the creation of worker nodes from within a Kubernetes cluster and creates InstanceGroup objects in your cluster. Additionally, instance-manager will provision the actual machines and bootstrap them to the cluster.
Worker nodes in Kubernetes clusters work best if provisioned and managed using a logical grouping. Kops introduced the term “InstanceGroup” for this logical grouping. In AWS, an InstanceGroup maps to an AutoScalingGroup.
Given a particular cluster, there should be a way to create, read, upgrade and delete worker nodes from within the cluster itself. This enables use-cases where worker nodes can be created in response to Kubernetes events, InstanceGroups can be automatically assigned to namespaces for multi-tenancy, etc.
instance-manager provides this Kubernetes native mechanism for CRUD operations on worker nodes.
You must first have atleast one instance group that was manually created, in order to host the instance-manager pod.
For installation instructions and more examples of usage, please refer to the Installation Reference Walkthrough.
$ kubectl create -f instance_group.yaml
instancegroup.instancemgr.keikoproj.io/hello-world created
$ kubectl get instancegroups
NAMESPACE NAME STATE MIN MAX GROUP NAME PROVISIONER STRATEGY LIFECYCLE AGE
instance-manager hello-world ReconcileModifying 3 6 hello-world eks crd normal 1msome time later, once the scaling groups are created
$ kubectl get instancegroups
NAMESPACE NAME STATE MIN MAX GROUP NAME PROVISIONER STRATEGY LIFECYCLE AGE
instance-manager hello-world Ready 3 6 hello-world eks crd normal 7mAt this point the new nodes should be joined as well
$ kubectl get nodes
NAME STATUS ROLES AGE VERSION
ip-10-10-10-10.us-west-2.compute.internal Ready system 2h v1.14.6-eks-5047ed
ip-10-10-10-20.us-west-2.compute.internal Ready hello-world 32s v1.14.6-eks-5047ed
ip-10-10-10-30.us-west-2.compute.internal Ready hello-world 32s v1.14.6-eks-5047ed
ip-10-10-10-40.us-west-2.compute.internal Ready hello-world 32s v1.14.6-eks-5047ed| Provisioner | Description | Documentation | API Reference | Maturity |
|---|---|---|---|---|
| eks | provision nodes on EKS | Documentation | API Reference | Production |
| eks-managed | provision managed node groups on EKS | Documentation | Experimental | |
| eks-fargate | provision a cluster to run pods on EKS Fargate | Documentation | Experimental |
To create an instance group, submit an InstanceGroup custom resource in your cluster, and the controller will provision and bootstrap it to your cluster, and allow you to modify it from within the cluster.
Please consider that this project is in alpha stages and breaking API changes may happen, we will do our best to not break backwards compatiblity without a deprecation period going further.
The previous eks-cf provisioner have been discontinued in favor of the Alpha-2 eks provisioner, which does not use cloudformation as a mechanism to provision the required resources.
In order to migrate instance-groups from versions <0.5.0, delete all instance groups, update the custom resource definition RBAC, and controller IAM role, and deploy new instance-groups with the new provisioner.
Please see CONTRIBUTING.md.
Please see DEVELOPER.md.